No. If we take −1/sqrtn,1/sqrtnn instead of 0,1n, the problem reduces to asking if we can have many unit vectors vj with pairwise scalar products −gamma or less where gamma>0 is a fixed number. But if we have N such vectors, then the square of the norm of their sum is at most N−gammaN(N−1). Since this square must be non-negative, we get N−1legamma−1 regardless of the dimension.
No comments:
Post a Comment